Beyond Vanity Metrics: Using Cohort Behavior to Identify Sustainable Growth

Understanding Sustainable Scaling Requires More Than Revenue Growth

The purpose of a startup is to discover and scale a repeatable, profitable business model. However, a startup cannot know if it’s scaling sustainably without accurately measured data. Understanding which growth mode the business operates in, based on how customers behave, is one of the most critical tools to know before deciding how to scale.

In Lean Analytics, Alistair Croll and Benjamin Yoskovitz identify three growth modes: Acquisition mode (focused on bringing in new customers), Hybrid mode (balancing acquisition and retention), and Loyalty mode (focused on maximizing value from existing customers). Early-stage e-commerce businesses, particularly those that have just gained traction, typically operate in Acquisition mode. In competitive markets with many alternatives, prioritizing customer acquisition is both rational and strategically necessary.

In the last blog series (refer to Why Metrics Lie: A D2C Case Study), transactional data was simulated for a D2C dog chew business. Although the online retailer has a successful social media ad campaign, the business faced two major obstacles to sustainable growth: very low repeat purchase rates and minimal referrals from existing customers.

Recap

In the blog Why Metrics Lie: A D2C Case Study, a simulated D2C dog chew brand appeared healthy on standard dashboards: revenue grew, paid social showed strong profits, and overall margins were positive. However, two critical patterns emerged:

• Low repeat purchase rates despite customers buying multiple units initially

• Dependency on paid acquisition rather than referrals

Standard dashboards revealed what was happening, but not why customers weren’t returning or whether growth was sustainable.

The organic acquisition channel was also performing well, indicating existing brand awareness and market demand. Based on this pattern, the intuitive response would be to differentiate the product to strengthen organic growth and increase referrals. However, this approach lacks critical information: which customers are most likely to refer others? Which segments show the highest repeat purchase rates? How long does it typically take for different customer types to make a second purchase?

Without understanding these customer-level patterns, pursuing product differentiation or organic growth strategies becomes speculation. Differentiation decisions depend heavily on which customer segments the business should prioritize, and aggregate data cannot reveal that.

Another critical question is which types of customers generate the most revenue for the business. At an aggregate level, revenue appears to be growing. As shown in the preceding blog, Why Metrics Lie: A D2C Case Study, the business ends the year with positive revenue and profit. The chart below further reinforces this pattern: both revenue and profit trend upward overall, despite a weak start in early 2026.

However, aggregate growth alone provides limited insight into the health of the business. Growth driven primarily by paid advertising is inherently unstable and exposes the business to unforeseen fluctuations in performance. The core business problem is therefore not whether revenue is currently growing, but whether that growth is controllable and sustainable. Blind faith in aggregate revenue trends offers no assurance: the business may grow another year, or it may stagnate and fail.

Sustainable scaling requires understanding which customers drive revenue and profit, why they do so, and how their behavior can be influenced over time. Without this insight, the business is reacting to outcomes rather than managing the underlying drivers of growth.

The Repurchase Rate: Why It Matters for the D2C Business

As discussed earlier, the business demonstrates a strong advertising strategy. Viewed in isolation, this suggests a clear focus on customer acquisition. The data supports this interpretation. To determine which growth mode the business is operating in, one would theoretically need to measure how many customers who purchased last year went on to purchase again this year. This is captured by the Annual Repurchase Rate (ARR), not to be confused with Annual Recurring Revenue (ARR).

However, the business has existed for only approximately twelve months, making a true annual repurchase calculation impossible. As a result, alternative approaches are required. One option is to analyze average purchase gap duration, which measures the average time it takes a customer to make a repeat purchase. Yet this metric alone is insufficient: it fails to account for customers who repurchase earlier or later than the average, and it ignores those who have not yet had the opportunity to repurchase.

Relying solely on repurchase gaps, therefore, risks recreating the very problem highlighted in Why Metrics Lie: A D2C Case Study: the use of vanity metrics that obscure underlying customer behavior rather than explain it. A more appropriate approach is to estimate the repurchase rate as a range, based on plausible repurchase windows, rather than a single point estimate. This allows the analysis to reflect uncertainty while remaining grounded in customer-level behavior.

The average time to repurchase is 45 days, while the maximum observed repurchase time is 60 days (approximately two months). To account for customers who repurchase later than the average, a time-based cohort structure is therefore best defined using a 60-day window, corresponding to the maximum observed repurchase interval.

Using this approach, the year is divided into six two-month periods. Customers are assigned to cohorts based on the period in which their first purchase occurs. The chart below visualizes these cohorts and tracks their repurchase behavior within each 60-day window.

The repurchase rate is calculated as the share of customers from the previous period who make a repeat purchase in the current two-month period. Formally, it is defined as the number of customers who reorder in the current period divided by the total number of customers in the previous period. This metric answers a specific question: What proportion of last period’s buyers returned to make another purchase?

The returning customer rate, by contrast, measures the share of customers in the current period who have purchased before. It is calculated as the number of returning customers divided by the total number of customers in the current period. This metric simply captures how much of the current demand is driven by repeat buyers.

Although both metrics relate to customer retention, they answer different questions. The repurchase rate is designed to assess whether customers who experienced product “wear-out” return to purchase again. A higher repurchase rate indicates stronger repeat-purchase behavior. The returning customer rate, on the other hand, reflects the composition of demand in a given period. While a higher level is generally preferable, the trend over time is more informative than the absolute value in any single period.

Examining the dashboard above, the returning customer rate shows a declining pattern. This implies that, in each period, new customers increasingly outnumber returning customers, an indicator of the business’s aggressive acquisition-driven growth strategy. This interpretation is reinforced by the rising new customer growth rate, which mirrors the decline in the returning customer share. At the same time, the repurchase rate remains relatively stable, though consistently below 30%, a level that is typical for a young e-commerce business.

One pattern, however, stands out across all charts: a sharp decline in revenue, customer acquisition, and repeat purchases, with returning customers nearly disappearing in the final period. One possible explanation is a breakdown or pause in the acquisition strategy. Another is product-market mismatch. The latter, however, is less likely given the presence of repeat purchases earlier, evidence of market demand, and periods of revenue and profit growth.

A more plausible explanation is that customer quality is not sustainable long term: referral rates remain low, repeat purchasing is weak, and limited social proof, such as bad reviews, may be hindering new customer acquisition. Regardless of the precise cause, the strategic implication is clear. The business must rethink its customer strategy to strengthen retention and referrals, enabling a transition from pure Acquisition mode toward a more sustainable Hybrid growth mode.

How Repurchase Rates Define Acquisition Cohorts

The previous section showed that the average repurchase time is approximately 45 days. This naturally raises the question: how do customers who repurchase earlier, later, or around this average actually behave over time? Rather than relying on a standard monthly or weekly cohort analysis, it is more informative to track customers based on when a repurchase is theoretically expected to occur.

The logic is straightforward. If the average repurchase cycle is 45 days, then customers who made their first purchase at time zero should, in theory, reappear in subsequent 45-day windows. Their performance can then be tracked across their lifetime within these expected repurchase intervals. At the same time, some customers may repurchase faster (e.g. within 30 days), while others may take longer (e.g. around 60 days). Capturing these dynamics requires cohort windows that reflect actual consumption and replacement behavior rather than arbitrary calendar months.

For this reason, retention was analyzed using three repurchase cadences: 30 days, 45 days, and 60 days.

The resulting cohort heatmaps reveal a striking and consistent pattern. Across both the 45-day and 60-day cohorts, customers make at most a single repurchase, after which retention collapses. In the 45-day cohort, for example, only 10.57% of customers who purchased in June 2025 returned during the subsequent 45–90 day window. Retention is similarly weak in the 60-day cohort, where at best roughly one-third of customers repurchase, and only once.

This outcome stands in sharp contrast to what would be theoretically expected for a consumable product. Under ideal conditions, a customer purchasing in January 2025 should, in the case of a 45-day repurchase cycle, make up to eight purchases per year (12 months ÷ 1.5 months), or approximately six purchases under a 60-day cycle. Instead, the data suggests that the product fails to generate a repeat habit, or at a minimum fails to stay relevant in the customer’s consideration set after the first repeat.

Several plausible explanations emerge. Dog owners may rotate across different types of chews, such as antler chews, nylon chews, or toys, rather than repeatedly purchasing the same product. In addition, the business currently offers only a medium-sized chew, which may not adequately match the diversity of dog sizes and bite strengths. Finally, the business appears to rely heavily on acquisition-driven growth. When customer acquisition costs are high, and repeat purchasing is weak, profitability can quickly deteriorate, even if revenue appears healthy in the short-run.

Taken together, these findings suggest that future marketing efforts should prioritize second-purchase conversion and post-purchase engagement rather than continued aggressive acquisition. Hence, retention, and not aggressive traffic.

Behavioral Cohorts

The 30-day, 45-day, and 60-day windows reveal more than what is obvious. They reveal distinct customer behaviors.

First are one-time buyers. These customers are largely acquired through aggressive acquisition channels. They respond to ads, make a purchase, but do not return. In other words, the ad is effective, but the product experience does not translate into repeat business.

Next are fast repeat buyers, who repurchase within 30 days. These customers may own dogs for whom a medium-sized chew is not suitable because the product wears out too quickly. Alternatively, they may have had a mixed first experience but, influenced by strong advertising, decided to give the product a second chance.

The third group revolves around the average repurchase window. For this product, a repurchase interval of 30 to 45 days is reasonable, depending on factors such as dog size and bite strength. A repeat purchase in this window suggests the product met expectations, as there is little reason for a satisfied customer not to return once the chew has been consumed.

Finally, moderate buyers repurchase between 46 and 60 days. This group likely includes owners of smaller dogs or customers who delay reordering for reasons that are not directly observable in the data. At this stage, these explanations remain hypotheses rather than firm conclusions. Nevertheless, classifying customers in this way already makes marketing decisions easier and more targeted.

Dividing customers by behavior allows the business to move beyond pure acquisition and toward a hybrid model, where customers are expected to repurchase more than once. When these behavioral cohorts are combined with the time-based cohorts introduced earlier (two-month periods based on observed repurchase limits), a much clearer picture of customer actions emerges.

Acquisition Channels by Behavioral and Time Cohorts

When acquisition channels are viewed through both behavioral and time cohorts, several clear patterns appear.

In the first acquisition period, there are no referrals, which is expected. More interestingly, referral activity declines over time for both the target and moderate cohorts, while remaining relatively stable for one-time buyers. In late 2024 (2024-P6), target buyers were often acquired through organic channel, indicating solid SEO performance for the brand. Although acquisition through this channel weakens in later periods. Across all channels, the marketing strategy of the startup appears to have a similar effect on one-time buyers, regardless of how they were acquired.

One notable exception is paid social. In 2025-P1, paid social acquisition increases for target buyers relative to other cohorts, suggesting that ads resonate more with customers already close to the expected repurchase window. However, the simultaneous decline in referrals indicates that expectations were not met.

In 2025-P2, customers who came through paid channels increased across cohorts, likely influenced by holiday seasons. While this drives volume, it does not materially improve repeat behavior.

Discounts, CAC, and the Discount Trap

Discount usage is highest among one-time buyers, despite their low retention. It is more stable among moderate buyers and fluctuates for target buyers. This indicates that discounts are often applied to customers least likely to stay, increasing Customer Acquisition Cost (CAC) without improving lifetime value.

For example, in 2025-P4, 58% of one-time buyers received a discount, yet this cohort contributed only 55% of lifetime revenue for that period. When the percentage of discounted customers exceeds their revenue contribution, it signals an inefficient discounting strategy. This pattern appears in multiple periods across cohorts: for target buyers in 2025-P3, 2025-P5, and 2025-P6; for one-time buyers in 2024-P6, 2025-P1, 2025-P4, and 2025-P6. In these periods, the cost to acquire customers, in this case, customers who received a discount, exceeds the revenue they generate.

Even though one-time buyers generate a large share of total revenue, they are expensive on a per-customer basis. This pattern aligns with what is commonly known as the discount trap: a business becomes reliant on frequent or deep discounts to drive sales, which gradually erodes margins, conditions customers to expect lower prices, attracts price-sensitive one-time buyers rather than loyal customers, and weakens the brand’s perceived value over time.

Strategic Recommendations

At first glance, revenue appears healthy. However, it is largely driven by one-time buyers acquired through paid channels. New customer growth is slowing, repurchase rates are low, and referrals are minimal. At the same time, strong organic traffic suggests that market demand exists. Hence, the problem is not a validation problem but of Churn.

The data (Please refer to the Data-Generating-Process) also shows that fulfillment issues are not strongly correlated with churn, ruling out operational failure as the primary cause.

The strategic objective should be to move from pure acquisition toward a hybrid growth model. While the current strategy effectively attracts first-time buyers, the business must convert a meaningful share of them into repeat customers.

There are several practical ways to improve repeat purchasing without relying solely on a subscription model. Subscriptions alone do not differentiate the product. Offering chews in different sizes or durability levels would allow customers to switch as their dog’s chewing behavior changes and better match product fit across cohorts. Adding a small range of complementary chews could keep customers within the brand rather than pushing them toward substitutes. Simple post-purchase reminders or usage guidance could help customers remember to reorder once the product is consumed. Targeted follow-ups or small incentives aimed specifically at the second purchase may be critical, as most drop-off occurs after the first repeat.

Once these behaviors are established, a subscription bundle could support repeat purchasing by making reordering easier, rather than relying on discounts or forcing long-term commitment. Clearer alignment between products and dog size or bite strength would further reduce one-off trials and improve repeat use.